Decision tree discrimination assisting device

Abstract

The present invention relates to a decision tree discrimination assistance device, which assists the improvement of analysis accuracy related to decision tree discrimination by ensuring the readability of an analysis process. A decision tree determination assistance device (1) is provided with: a setting unit (13) that sets a decision tree on the basis of a threshold value that uses a variable, and sets a prediction result of the suitability or insuitability of a branch determined by the decision tree; a precision calculation unit (14) that calculates the analysis precision of the decision tree, said analysis precision being composed of the suitability rate and the reproduction rate, on the basis of the suitability data of the corresponding data group and the prediction result; a recording unit (15) that accumulates and records the layering conditions and analysis accuracy of the branches having the most number of analysis data predicted to be appropriate among the branches of each decision tree every time the settings related to the setting unit (13) are changed; and an output unit (16) that causes a display device to output the recorded layering conditions of the branches and the analysis accuracy.

Description

Technical Field

[0001] This invention relates to a decision tree discrimination aid device for assisting in decision tree-based discrimination analysis. Background Technology

[0002] As such a technology, for example, Patent Document 1 proposes a multiple regression analysis device that outputs information about the consistency with measured values ​​in the multiple regression equation and information about the consistency with measured values ​​in the combined multiple regression equation to the user.

[0003] Patent Document 1: Japanese Patent Application Publication No. 2022-064707

[0004] However, the analytical apparatus described in Patent Document 1 is intended for studying the logic and visual design of improving the accuracy of linear multiple regression analysis through hierarchical analysis. Therefore, in order to improve the accuracy of decision tree-based analysis, it is important for the user to understand the changes in the thresholds of the variables involved in the decision tree and the changes in the accuracy of the analysis, which cannot ensure the readability of the analysis process. Summary of the Invention

[0005] The present invention has been made in view of this purpose, and its object is to provide a decision tree discrimination aid device that assists in improving the accuracy of decision tree-based discrimination by ensuring the readability of the analysis process.

[0006] In view of the above-mentioned problems, the decision tree discrimination assistance device according to the present invention assists in the analysis based on decision tree discrimination for all data sets having multiple analysis data, wherein the analysis data includes attribute data related to attributes, values ​​of multiple adopted variables, and appropriateness or inappropriateness data as the result of discrimination. The decision tree discrimination assistance device is characterized by comprising: a storage unit that stores the analysis data of all the data sets; an extraction unit that extracts the corresponding data set from the stored data sets and the corresponding attribute data selected by the input device; and a setting unit that, via the input device, sets the corresponding data set according to the values ​​of the adopted variables and the threshold values ​​of the adopted variables that become stratification conditions. A decision tree is set according to the data set, and a prediction result is set to determine whether a branch determined by the decision tree is appropriate or inappropriate; an accuracy calculation unit calculates the analysis accuracy of the decision tree, consisting of fit rate and reproduction rate, based on the appropriateness data of the corresponding data set and the prediction result set by the setting unit; a recording unit accumulates and records the stratification conditions and analysis accuracy of the branch with the most analysis data predicted as appropriate in each branch of the decision tree each time the settings involved in the setting unit are changed; and an output unit causes the display device to output the recorded stratification conditions and analysis accuracy of the branch.

[0007] According to the present invention, each time the settings involved in the setting unit are changed, the content obtained by accumulating and recording the hierarchical conditions and analysis accuracy of the branch with the most analysis data predicted as appropriate in each branch of the decision tree can be displayed on the display device. As a result, by ensuring the readability of the analysis process, it is possible to help improve the analysis accuracy based on decision tree discrimination. Attached Figure Description

[0008] Figure 1 (a) is a block diagram of a decision tree discrimination aid device. Figure 1 (b) is a diagram representing an example of all the data sets in the analysis.

[0009] Figure 2 It means by Figure 1 The diagram shows an example of a decision tree and analysis results displayed on the display device of the decision tree discrimination aid device.

[0010] Figure 3 It means by Figure 1 The diagram shows the recorded content of the recording section displayed on the display device of the decision tree discrimination auxiliary device.

[0011] Explanation of reference numerals in the attached figures

[0012] 1… Decision tree discrimination auxiliary device; 11… Storage unit; 12… Retrieval unit; 13… Setting unit; 14… Accuracy calculation unit; 15… Recording unit; 16… Output unit; 20… Input device; 30… Display device. Detailed Implementation

[0013] The following is for reference Figures 1-3 The decision tree discrimination aid device according to embodiments of the present invention will be described. For example... Figure 1 As shown, the decision tree discrimination assist device 1 according to this embodiment is a device that assists in the analysis based on decision tree discrimination of all data sets having multiple analysis data. The analysis data includes attribute data related to one of the multiple attributes, values ​​of multiple adopted variables that are associated with the attribute data, and appropriateness or inappropriateness data as the result of the discrimination. For example, it is a device that assists users in using manufacturing conditions such as the composition ratio of materials such as metal materials or resin materials and the heating time during manufacturing as adopted variables and analyzing the evaluation results (e.g., tensile strength, surface hardness, etc.) of the manufactured materials through decision tree discrimination.

[0014] A processing device 10 is provided to assist in decision tree-based analysis of all data sets described later. An input device 20 and a display device 30 are connected to the processing device 10. The processing device 10 has a storage device (not shown) for storing all data sets and setting conditions, and a processing device (not shown) for calculating analysis accuracy, etc., as hardware. The processing device 10 has a storage unit 11, an extraction unit 12, a setting unit 13, an accuracy calculation unit 14, a recording unit 15, and an output unit 16 as software.

[0015] Storage unit 11 stores the analysis data for all data sets. For example... Figure 1 As shown in (b), the analytical data is each of data 1 to data N, and the total data set is all the data of data 1 to data N. Each analytical data set has attribute data related to attributes 1 and 2. In this embodiment, attribute 1 includes A and B, and attribute 2 includes a to c. The attributes mentioned here can be, for example, the part name of the material that is the object of the analytical data, the usage environment, the general name of the material, etc. In this embodiment, the values ​​of attribute data X1 to XN, Y1 to YN, and Z1 to ZN are respectively associated with the variables X, Y, and Z of data 1 to data N, which are the analytical data. For example, the variables X, Y, and Z are variables such as the content of a specific component of the material during the manufacturing process, the heating temperature, and the heating time. Furthermore, data 1 to data N, which are the analytical data, are associated with either "appropriate" or "inappropriate" as a result of the judgment, which is used as appropriateness data. It is also possible to set "appropriate" to 1 and "inappropriate" to 0 to establish a correspondence. For example, if the tensile strength or surface hardness of the material analyzed is above the desired value, it is set as appropriate; if it does not meet the desired value, it is set as inappropriate.

[0016] The extraction unit 12 extracts the corresponding data group from all stored data groups, which corresponds to the attribute data selected via the input device 20. For example, such as Figure 2 As shown, the user selected attribute data "A" in attribute 1 and attribute data "b" in attribute 2. In this case, from Figure 1 Extract 100 data points from all the data sets shown in (b) as the corresponding data sets.

[0017] The setting unit 13, via the input device 20, sets a decision tree for the corresponding data set based on the adopted variables and the threshold values ​​of the adopted variables that become stratification conditions, and sets the prediction result of whether the branch determined by the decision tree is appropriate or inappropriate. For example, such as Figure 2As shown, a decision tree is set up with thresholds XA, XB, YA, YB, ZA, and ZB for variables X, Y, and Z, indicating that the predicted result is appropriate. The prediction results for the analysis data at the end of each branch are also set. Here, there are four branches that are judged as "appropriate", and the number of analysis data that meet the criteria in each branch is "4", "13", "4", and "1" from left to right.

[0018] The accuracy calculation unit 14 calculates the analysis accuracy of the decision tree, consisting of the fitness rate and the reproducibility rate, based on the appropriateness data of the corresponding data set and the set prediction results. Specifically, the number of stored appropriateness data set points set as appropriate is 40, and the number of inappropriateness data set points is 60. On the other hand, in the prediction results made by the user through the decision tree, the number of appropriateness data set points is 22, and the number of inappropriateness data set points is 78. The accuracy calculation unit 14 creates an analysis based on the appropriateness data of the corresponding data set and the set prediction results. Figure 2 The confusion matrix shown in the lower left corner is used to calculate the fitness and reproducibility of the decision tree.

[0019] Each time the settings involved in the setting unit 13 are changed, the recording unit 15 accumulates and records the hierarchical conditions and analysis accuracy of the branch (main flow) in each decision tree branch that has the most analysis data predicted as appropriate. Specifically, Figure 2 The main flow of the decision tree branches conforms to this. For example, these results are as follows: Figure 3 Accumulate records as shown. Figure 2 The results shown were recorded in Figure 3 The fourth row of the table shown is designated as setting condition 1. Other changed setting conditions and their results are also recorded in other rows. Output unit 16 causes display device 30 to... Figure 2 and Figure 3 The method shown outputs the stratification conditions and analysis precision of the recorded branches.

[0020] In this way, whenever the settings involved in the setting unit 13 are changed, the content obtained by accumulating and recording the stratification conditions and analysis accuracy of the branch with the most analysis data that is predicted to be appropriate in each branch of the decision tree can be displayed on the display device 30. As a result, users can easily improve the accuracy of discriminant analysis before and after stratification and make efficient analysis judgments by tracking variable changes.

[0021] The embodiments of the present invention have been described in detail above, but the present invention is not limited to the above embodiments, and various design changes can be made without departing from the spirit of the present invention described in the technical solution.

Claims

1. A decision tree discrimination aid device for assisting in the analysis of all data sets having multiple analytical data based on decision tree discrimination, wherein the analytical data includes attribute data related to attributes, values ​​of multiple adopted variables, and appropriateness or inappropriateness data as the result of the discrimination. The decision tree discrimination auxiliary device is characterized by having: The storage unit stores the analysis data for all the data sets mentioned above; The extraction unit extracts the corresponding data group from all the stored data groups, which corresponds to the attribute data selected via the input device. The setting unit, via the input device, sets a decision tree for the corresponding data set based on the value of the adopted variable and the threshold of the adopted variable that becomes the stratification condition, and sets a prediction result of whether the branch determined by the decision tree is appropriate or inappropriate. The accuracy calculation unit calculates the analysis accuracy of the decision tree, which consists of the fitness rate and the reproducibility rate, based on the appropriateness data of the corresponding data group and the prediction result set by the setting unit. The recording unit accumulates and records the hierarchical conditions and the analysis accuracy of the branch with the most analytical data that is predicted to be appropriate in each branch of the decision tree whenever the settings involved in the setting unit are changed. as well as The output unit enables the display device to output the recorded layering conditions of the branch and the analysis accuracy.

Images